The CAR-HEMATOTOX model for CAR T-cell hematotoxicity in R/R LBCL

Hematotocixity is a frequent adverse event related to chimeric antigen receptor (CAR) T-cell therapy. Despite an improvement in response rates in relapsed/refractory (R/R) B-cell malignancies, the utility of CAR T-cell therapy has been affected by its distinct toxicity profile.¹ CAR T-cell-mediated hematotoxicity is a manifestation of cytokine release syndrome and a component of CAR T-cell-associated toxicity, which can predispose patients to infectious complications.² There is a high incidence of Grade ≥3 neutropenia (30–38%), thrombocytopenia (21–26%), and anemia (5–17%) which persists after Day 21, highlighting the importance of hematotoxicity in the first year after CAR T-cell therapy. The underlying mechanism is poorly understood.^1,2

Rejeski, et al. recently reported the results from their study in which they devised and validated a risk stratification tool (CAR-HEMATOTOX) to identify patients who are at risk of hematotoxicity. Rejeski presented the study data at the 16th International Conference on Malignant Lymphoma (16-ICML)² and has also recently published the work in Blood.¹

Study design and patient characteristics^1,2

A multicenter retrospective analysis of 258 patients (intent-to-treat population) to assess patterns of hematopoietic reconstitution and to identify predictive markers. Patients were receiving either axicabtagene ciloleucel (n = 170) or tisagenlecleucel (n = 88) immunotherapy for R/R large B-cell lymphoma and were split into training and validation cohorts (see Figure 1).

Figure 1. Number of patients included in the training and validation cohorts* 

*Adapted from Rejeski, et al.¹

The primary endpoint was clinically significant neutropenia (absolute neutrophil count [ANC] ≤500/µL) between Days 0 and 60.

At the time of lymphodepletion, there were differences between the US and European cohorts; the European cohort had a higher tumor burden, more inflammation, and decreased ANC (see Table 1). 

Table 1. Patient characteristics at baseline*

Patients across all three cohorts had a high incidence of hematotoxicity; 62% had severe thrombocytopenia, 69% had anemia, 91% had severe neutropenia, and 64% had prolonged neutropenia (≥21 days after CAR transfusion). The median duration of severe neutropenia was 9 days (95% confidence interval [CI]; 8–10 days).

Clinical phenotypes of neutrophil recovery^1,2

The team found that neutropenia generally follows a bimodal curve, with recovery after G-CSF stimulation, followed by a second ‘dip’ in ANC.

Different patterns of recovery were observed, and patients in the training and European validation cohort could be divided into the following three groups:

• Quick recovery: sustained neutrophil recovery without a second dip in ANC <1,000/µL (n = 37)
• Intermittent recovery: neutrophil recovery with ANC above 1500 per µL, with a second dip of ANC to <1,000/µL (n = 78)
• Aplastic: continuous severe neutropenia with ANC <500/µL for ≥14 days (n = 34)

The pattern of thrombocytopenic depletion and recovery was quite different, with the lowest platelet count occurring between Days 21 and 60 following lymphodepletion. Platelet recovery is achieved by most patients around Day 90. This pattern suggests a different mechanism behind thrombocytopenia.

Biomarkers of prolonged neutropenia¹

Univariate analysis of pre-lymphodepletion biomarkers identified markers associated with a low hematopoietic reserve (low baseline platelet count, hemoglobin, or ANC) or increased inflammation (elevated C-reactive protein) were significantly associated with prolonged duration of neutropenia.

CAR-HEMATOTOX¹

A discriminatory multivariate model that could detect the binary outcome of severe neutropenia for <14 days or ≥14 days was developed using markers identified with an area under the curve (AUC) >0.6, a p value ≤0.1, and an odds ratio ≥2.5. Multivariate analysis of markers included in the model identified platelet count and ferritin as of particular importance, so they were weighted accordingly.

The CAR-HEMATOTOX model gives a score based on patient baseline features as described in Table 2. A final score of 0–1 is classed as low for severe neutropenia lasting ≥14 days, and a score ≥2 would be classed as high. In the training cohort, CAR-HEMATOTOX could discriminate for severe neutropenia lasting ≥14 days (AUC, 0.82; p < 0.001; sensitivity, 0.96; specificity, 0.67).

Table 2. CAR-HEMATOTOX model scoring based on baseline features*

CAR-HEMATOTOX was then tested in the two validation cohorts (European and USA), which were pooled together (n = 180). Using CAR-HEMATOTOX to divide patients into low- and high-score categories, 81% of the high-score group suffered with prolonged neutropenia (≥21 days after CAR transfusion), but only 42% of the low-score group had prolonged neutropenia (Table 3).

The primary endpoint of clinically significant neutropenia (ANC ≤500/µL) between Days 0 and 60 was seen for a total of 5.5 days (95% CI, 5–8) in the low-score group and 12 days (95% CI, 10–16) in the high-score group (Table 3), demonstrating the high negative predictive value of the score.

Table 3. Patients in CAR-HEMATOTOX high-/low-risk categories and relationship with outcomes*

Linear regression analysis of the two validation cohorts confirmed the discriminatory capacity of the CAR-HEMATOTOX score (USA cohort: AUC, 0.91; p < 0.001; European cohort: AUC, 0.77; p < 0.001). In terms of the clinical phenotypes of neutrophil recovery, most aplastic phenotypes fell within the high score group, whereas most of the quick recovery phenotypes were within the low score group. The study group also found that CAR-HEMATOTOX score was indicative of hospital stay, with those in the high score group having a longer hospitalization.

Conclusion

The authors concluded that real-world experience demonstrated the high incidence of delayed cytopenias following CAR T-cell therapy. They found that patients have different patterns of neutrophil recovery and can be separated into three groups (quick, intermittent, and aplastic).² The CAR-HEMATOTOX model was reported to be easy to apply and demonstrates the importance of pre-CAR T-cell therapy bone marrow reserve and inflammation as key features affecting cytopenia.^1,2 The study had limitations, in that it was retrospective, and had incomplete data around bone marrow infiltration and presence of dysplastic changes or clonal hematopoiesis. In addition, the team chose a lower cut-off score for the CAR-HEMATOTOX model, which yielded a highly sensitive tool with good negative predictive value; however, in the clinical setting, a higher cut-off may be necessary to improve specificity and positive predictive value.¹ The group concluded that CAR-HEMATOTOX could be used for risk-stratification prior to treatment and could help guide prophylactic treatment and outpatient management.^1,2